interesting and merit consideration, this secondary reviewer is not convinced that theassociation should be used to provide advice to practitioners in labeling.

7.3.3. Secondary Endpoints

Results from the 2° endpoints are shown in Table 9. The triple composite endpoint wasstatistically significant in favor of prasugrel at Days 30 and 90. (Although these were denotedas 2° endpoints, they are, in fact, sensitivity analyses on the 1° endpoint.)

The other 2° endpoints were statistically significantly in favor of prasugrel for the All ACSpopulation, and to lesser extents, for the UAINSTEMI and STEMI populations individually.

The stent thrombosis endpoint is robust (0.49 RR in favor of prasugrel, 95% CI 0.36, 0.68, forthe overall ACS population, p<0.001). Initially, the clinical reviewer (Dr. Karen Hicks) raisedconcerns regarding the validity of the stent thrombosis endpoint, because the CEC review didnot meet the diagnostic standards for stent thrombosis developed recently by the AcademicResearch Consortium (2007). These standards require angiographic confirmation of stentthrombosis, generally determined by an angiographic core laboratory or pathologicalconfirmation: evidence of recent thrombus within the stent or direct examination of tissueretrieved following thrombectomy. In TAAL, there was no review of angiograms by anangiographic core laboratory, and there was limited pathological confirmation; only reports ofcoronary angiograms and other clinical reports were use to make determinations of stent

Table 9: TAAL - Secondary Endpoints

Clopidogrelendpoint

Patientpopulation

N

Prasugrel

n % N n % N

Total

n %

CoxProportional HR

(95% C.I.)p

Composite of CV death, nonfatal MI, or UTVR at Day 30UAINSTEMI 5044 281 5.57 5030 349 6.94 10074 630 6.25 0.80 (0.68, 0.93) 0.005

STEMI 1769 118 6.67 1765 155 8.78 3534 273 7.72 0.75 (0.59, 0.96) 0.02AIIACS 6813 399 5.86 6795 504 7.42 13608 903 6.64 0.78 (0.69, 0.89) <0.001

Composite triple endpoint at Day 30UAINSTEMI 5044 274 5.43 5030 336 6.68 10074 610 6.06 0.81 (0.69, 0.95) 0.009

STEMI 1769 115 6.50 1765 166 9.41 3534 281 7.95 0.68 (0.54, 0.87) 0.002AIIACS 6813 389 5.71 6795 502 7.39 13608 891 6.55 0.77 (0.67, 0.88) <0.001

Composite of CV death, nonfatal MI, or UTVR at Day 90UAINSTEMI 5044 345 6.84 5030 420 8.35 10074 765 7.59 0.81 (0.70,0.94) 0.004

STEMI 1769 127 7.18 1765 168 9.52 3534 295 8.35 0.75 (0.59, 0.94) 0.013AIIACS 6813 472 6.93 6795 588 8.65 13608 1060 7.79 0.79 (0.70, 0.90) <0.001

Composite triple endpoint at Day 90UAINSTEMI 5044 333 6.60 5030 395 7.85 10074 728 7.23 0.83 (0.72, 0.97) 0.015

STEMI 1769 129 7.29 1765 178 10.08 3534 307 8.69 0.72 (0.57, 0.90) 0.004AIIACS 6813 462 6.78 6795 573 8.43 13608 1035 7.61 0.80 (0.71, 0.90) <0.001

Composite triple endpoint or re-hospitalization for cardiac ischemic eventsUAINSTEMI 5044 598 11.86 5030 688 13.68 10074 1286 12.77 0.86 (0.77, 0.96) 0.006

STEMI 1769 199 11.25 1765 250 14.16 3534 449 12.71 0.78 (0.65,0.94) 0.009AIIACS 6813 797 11.70 6795 938 13.80 13608 1735 12.75 0.84 (0.76, 0.92) <0.001

Composite of all-cause mortality, nonfatal MI, or nonfatal strokeUAINSTEMI 5044 504 9.99 5030 590 11.73 10074 1094 10.86 0.84 (0.75, 0.95) 0.005

STEMI 1769 188 10.63 1765 232 13.14 3534 420 11.88 0.80 (0.66, 0.97) 0.02AIIACS 6813 692 10.16 6795 822 12.10 13608 1514 11.13 0.83 (0.75, 0.92) <0.001

Definite or probable stent thrombosis per Academic Research Consortium (ARC) definition at study endUAINSTEMI 4798 39 0.81 4789 80 1.67 9587 119 1.24 0.49 (0.34, 0.72) <0.001

STEMI 1624 19 1.17 1633 40 2.45 3257 59 1.81 0.50 (0.29, 0.87) 0.011AIIACS 6422 58 0.90 6422 120 1.87 12844 178 1.39 0.49 (0.36, 0.68) <0.001

Prasugrel Secondary Review, page 43 of77

thrombosis.

The sponsor argued (regulatory response of August 22, 2008) that according to FDA draftguidance, an angiographic core laboratory is not required: "FDA strongly recommends thatinterpretation of data from tests such as angiograms, IVUS, and ECGs be performed byindependent core labs and that blinded adjudication of clinical events be conducted by a clinicalevents committee (CEC Clinical adjudication committees should be independent of core labanalysis centers to avoid potential bias).,,3

Ultimately, Dr. Hicks selected a number of cases for review by an independent core laboratory,and requested details regarding the adjudication process. The independent review appeared tosupport the reliability of the original results.

7.3.4. Efficacy Conclusions

Treatment with prasugrel was associated with a statistically significant reduction in thecomposite triple endpoint of cardiovascular death, nonfatal MI, and nonfatal stroke. Thesefindings were statistically persuasive across the UAINSTEMI population, the STEMI population,and the overall ACS population, and robust to exploration. The effect of prasugrel on the 10

endpoint was evident across the spectrum of subject weight, age, and sex, and in the presenceand absence of concomitant diseases and medications that are common in the ACS population.Results were similar whether or not subjects received a stent, and irrespective of whether abare metal stent or drug-eluting stent was deployed.

Efficacy was driven by a reduction in non-fatal MI, which was statistically significant in both theSTEMI and UAINSTEMI populations. There was a positive trend in mortality in favor ofprasugrel in the STEMI population, but not in the larger UAINSTEMI population. Stroke wassimilar in the two groups. In exploratory analyses, variability in salt to base conversion had nodemonstrable effect on prasugrel's efficacy.

The following weaknesses and concerns have been identified:

1) Prevention of stroke: Importantly, the efficacy of c1opidogrel was established in CURE,where clopidogrel was compared to placebo on a background of aspirin in subjects presenting .with UAINSTEMI. The study utilized a triple composite endpoint similar to that used in TAAL. InCURE, c1opidogrel was associated with a 20% relative risk reduction on the triple endpoint, butwas essentially neutral on the stroke component of the endpoint. Specifically, rates of strokewere 1.2% and 1.4% for the c1opidogrel and placebo groups, respectively, for a non-statisticallysignificant relative risk reduction of 14% (95% C.1. -17.7% to 36.6%). In TAAL, prasugrel'seffect on stroke was neutral with respect to clopidogrel (hazard ratio 1.02 in favor of clopidogrel,95% C.1. 0.71 to 1.45). Therefore, in estimating what prasugrel's effect on stroke would havebeen relative to placebo, the neutral effects in CURE and TAAL are chained, and the evidenceof effectiveness is nil.

2) For subjects with a prior history of TIA or stroke, the overall effect of prasugrel wasnegative, driven by a striking increase in strokes (hazard ratio of 5.64,95% C.1. 1.65 to 19.3).(Of note, subjects with a history of hemorrhagic stroke were excluded from participation, and itis possible that inclusion of such patients might have driven the risk of recurrent stroke evenhigher.) Presently, the evidence that prasugrel causes stroke in patients with a prior TIA or

3Guidance for Industry: "Coronary Drug-Eluting Stents-Nonclinical and Clinical Studies," draft, March2008. http://www.fda.gov/cdrh/ode/guidance/6255.html

Prasugrel Secondary Review, page 44 of77

stroke seems more persuasive than the evidence that prasugrel prevents stroke in those withoutsuch a history. As such, it would not be appropriate to give prasugrel an indication for stroke,based on extant data. On the contrary, risk management should include a contraindication forpatients with a prior history of TIA or stroke.

3) Subjects of African descent: Subjects of African descent accounted for less than 3% ofthe subject population in TAAL. At this point, there is no reason to believe that results fromCaucasians can not be extrapolated to patients of African descent, but the size of the subgroupwas too limited to be very informative in its own right.

7.4. Safety

7.4.1. Exposure

TALL included 6741 subjects in the prasugrel treated population and 6716 subjects in theclopidogrel treated population (13,457 in total). Taking into consideration temporary drugdiscontinuations, median exposure was 442 days in the prasugrel group and 444 days in theclopidogrel group. Over 4200 subjects in each treatment group were exposed for greater thanone year.

Although TAAL was a large cardiovascular outcome study, it was by no means a large "simple"trial. Subjects were evaluated at hospital discharge, Days 30, 90, 180, 270, 360, and 450 (orlast visit) for adverse events and concomitant medications. In addition, vital signs, ECG,complete blood count, platelet count, and clinical chemistries were performed at each visit.Thus, the safety database is quite robust.

Because 98.8% of randomized subjects received the study agent, the safety population is notimportantly different from the ITT efficacy population. As such, the reader is referred back toTable 2 and Table 3 for a breakdown of demographic and historical characteristics, respectively.

The following weaknesses are identifiable in terms of exposure: the database included fewsubjects with hepatic and renal impairment. Approximately 0.5% of subjects in each group hadpre-existing hepatic impairment; approximately 0.8% had severe renal impairment (calculatedcreatinine clearance < 30 mLlmin). Approximately 10% of subjects had calculated creatinineclearance between 30-60 mLlmin. Thus, experience is extremely limited in subjects with severehepatic and renal dysfunction, and this should be pointed out in labeling.

7.4.2. All-Cause Mortality

Table 10 displays the sponsor's summary breakdown of deaths in TAAL, adapted from TableTAAL.11.1 0 of the TAAL study report. The right-most column provides point estimates for thenumbers of events that prasugrel would be expected to prevent (if >0) or cause (if <0), relativeto clopidogrel, per 1000 patients treated.

There was no significant difference in all-cause death between treatment groups; thefrequencies of CEC-adjudicated all-cause mortality were 2.76% and 2.90% in the prasugrel andc1opidogrel treatment groups, respectively (p=0.64, Table 10). Differences in mortality in thevarious categories are not statistically significant, but the most favorable trends for prasugrel(fewer deaths) are in those classified as related to acute MI and sudden/unwitnessed. The mostunfavorable trends for prasugrel are in deaths classified as hemorrhagic/non-ICH, ICH, andmalignancy.

Prasugrel Secondary Review, page 45 of77

Deaths due to bleeding and malignancy are addressed more fully in sections below.

Table 10: Summary of Deaths in TAAL (adapter from sponsor's Table TAAL.11.10)delta events per

1000 patientstreated (positive =

Prasugrel Clopidogrel favorable forn=6813 n=6795 prasugrel)

n % n %

All Cause Death 188 2.76 197 2.9 1.4

Cardiovascular (conponent of 10 efficacy endpoint) 133 1.95 150 2.21 2.6atherosclerotic vascular disease (excluding coronary) 0 0 3 0.04 0.4CHF/cardiogenic shock 31 0.46 30 0.44 -0.1related to CABG or PCI 15 0.22 16 0.24 0.2dysrhythmia 4 0.06 7 0.1 0.4pulmonary embolism 3 0.04 0 0 -0.4acute MI 24 0.35 36 0.53 1.8sudden or unwitnessed 36 0.53 42 0.62 0.9ICH 9 0.13 5 0.07 -0.6non-hemorrhagic stroke 5 0.07 6 0.09 0.1other cardiovascular 6 0.09 5 0.07 -0.1

Non-Cardiovascular 55 0.81 47 0.69 -1.2accident/trauma 4 0.06 4 0.06 0.0hemorrhage, non-ICH 9 0.13 1 0.01 -1.2infection 11 0.16 10 0.15 -0.1malignancy 21 0.31 17 0.25 -0.6suicide 3 0.04 2 0.03 -0.1other 7 0.1 13 0.19 0.9

7.4.3. Discontinuations

The most commonly cited reason given for discontinuation was "subject decision," reported inapproximately 9% of subjects in each treatment group. The second most common reason fordiscontinuation was an adverse event, with 7.2% and 6.3% of subjects discontinuing in theprasugrel and c1opidogrel groups, respectively (Table TAAL 12.2, TAAL Clinical Study Report).Hemorrhagic adverse events accounted for essentially all of the disparity: the percentages ofsubjects discontinuing study drug due to a serious hemorrhagic event were 1.6% and 0.9% inthe prasugrel and c1opidogrel groups, respectively. For non-serious hemorrhagic events, therespective percentages were 0.9% and 0.5%. The numbers of discontinuations for non­hemorrhagic adverse events were similar in the two groups.

7.4.4. Intracranial Hemorrhage (ICH)

In TAAL, ICH was reported in 20 (0.29%) and 16 (0.24%) subjects in the prasugrel andclopidogrel groups, respectively. In both groups, the majority of events occurred between 30and 180 days post-randomization. Intracranial hemorrhages in the prasugrel group were moresevere and recovery from these events was lower than in the c1opidogrel group. Compared toclopidogrel, twice as many prasugrel-treated subjects died from ICH.

Prasugrel Secondary Review, page 46 of77

7.4.5. Non-ICH Bleeding

The sponsor categorized bleeding events as related or unrelated to coronary artery bypass graft(CABG) surgery. Events within 7 days of completion of the CABG surgery were classified asCABG-related by the central adjudication committee.

7.4.6. Non-CABG-Related Bleeding

The risk of bleeding was well-considered in the review by Dr. Hicks. Prasugrel was associatedwith excess bleeding relative to c1opidogrel, irrespective of bleeding definition, seriousness, orlocation, and across most subgroups assessed. The time course of CEC-adjudicated TIMImajor or minor bleeding is shown Figure 15. Note that approximately one-third of all bleedingevents were recorded in the first day; nearly half of all bleeding events were reported in theinitial 10 days.

Figure 15: Incidence of Non-CABG-Related TIMI Major or TIMI MinorBleeding Events - All ACS Population

•• _. clopidogrel ., .,_ •••••••••.,..'_.,, __ fI

I'..,~.... #.',r· ,.",.'.-

- prasugrel

6

5

--cQ)

4>Q)

..c--·3 3--cQ)u 2'-Q)0.

oo 3 6 9 10 120 230 340 450

time (days)

Table 11 summarizes the various categories of bleeding events in TAAL. Because somesubjects experienced more than one bleeding event, they appear in more than one category.The last two categories of the upper section, "Worst: TIMI Minor" and "Worst: TIMI Minimal,"represent the subjects in whom the most significant bleeding event was a TIMI minor or TIMIminimal bleeding event, respectively.

There were 21 and 5 fatal bleeding events in the prasugrel and clopidogrel groups, respectively(RR = 4.19,95% C.I.: 1.58, 11.1, p=O.002), Table 11. All 5 of the fatal bleeding events in thec1opidogrel group were intracranial in location. For the prasugrel group, 9 of 21 fatal bleedingevents were intracranial, and 12 were not (5 were gastrointestinal [GI], 2 originated frompuncture sites, 2 from surgical sites, 2 from retroperitoneal locations, and 1 from an intra­abdominal location). Given that it is generally more feasible to manage bleeding at extra-cranialsites than at intracranial sites, it is worth emphasizing that none of the deaths in the clopidogrelgroup, but over half the deaths in the prasugrel group, were attributed to extra-cranial sites of

Prasugrel Secondary Review, page 47 of77

hemorrhage. The disparity in deaths from extracranial hemorrhage between the prasugrel andclopidogrel groups suggests that severe bleeding may be more difficult manage in patients whoreceived prasugrel.

The RR was 1.52 for TIMI life-threatening bleeding events, and this was also statisticallysignificant (Table 11). For TIMI major and TIMI minor bleeding, the relative risks were 1.32 and1.31, respectively, and the differences were statistically significant.

From these data, it is possible to characterize bleeding in terms of excess bleeding events per1000 patients treated. Comparing prasugrel to clopidogrel, the absolute risks predict 2.4additional fatal bleeding events, 4.3 additional TIMI life-threatening bleeds, 5.1 additional TIMImajor bleeds (which include fatal and life-threatening bleeds), 5.4 additional TIMI minor bleeds,and 19.4 additional TIMI minimal bleeds per 1000 patients treated. In total, per 1000 patientstreated, these calculate to 30 excess TIMI bleeding events of any magnitude, 10.5 bleedingevents associated with a decrease in hemoglobin of ~ 3 g/dL, and 5.1 bleeding eventsassociated with a decrease in hemoglobin of ~ 5 g/dL.

7.4.7. CABG-Related Bleeding

The prasugrel-associated bleeding risk was particularly malignant in subjects who underwentCABG (Table 11, bottom). In the prasugrel group, there were 24 TIMI major bleeding events in213 total ACS subjects (11.3%, RR=3.50), of which 2 were fatal (0.9%). In the c1opidogrelgroup, there were 8 TIMI major bleeds, and none were fatal. There are additional analyses ofCABG-related bleeding on page 43.

Reviewer's Comments: Prasugrel should not be the drug of choice for patients in whom CABO surgery isanticipated. From a practical standpoint, prasugrel is not well-suited for pre-treatment of patients inwhom coronary anatomy is unknown.

CDER undertook independent analyses of bleeding adverse events, characterized as "mild,""moderate," or "severe," as well as those meeting the regulatory definition of a serious adverseevent (see primary clinical review). For all categories of bleeding events, the RR wasapproximately 1.4, and the difference between treatment groups was statistically significant.The frequencies of bleeding events meeting the regulatory definition of a serious adverse eventwere 5.5 and 3.8% in the prasugrel and c1opidogrel groups, respectively (RR 1.46, 95% C.1.1.25,1.71).

Prasugrel Secondary Review, page 48 of 77

Table 11: CEC Adjudicated Bleeding

Non-CABG-Related

bleeding endpoint Prasugrel Clopidogrel HR (95% C.I.) p

N n % N n %

TIMI Fatal 6741 21 0.3 6716 5 0.1 4.19 (1.58,11.1) 0.002

TIMI Life-Threatening 6741 85 1.3 6716 56 0.8 1.52 (1.08,2.13) 0.015

TIMI Major 6741 146 2.2 6716 111 1.7 1.32 (1.03,1.68) 0.029

TIMI Minor 6741 164 2.4 6716 125 1.9 1.31 (1.04,1.66) 0.022

TIMI Minimal 6741 460 6.8 6716 314 4.7 1.47 (1.28,1.70) 0.022

CABG-Related

bleeding endpoint Prasugrel Clopidogrel HR (95% C.I.) P

N n % N n %

TIMI Fatal 213 2 0.9 224 0 0.0

TIMI Major 213 24 11.3 224 8 3.6 3.50 (1.53,7.99) 0.002

The fatality rate for intracranial hemorrhages was twice as high in the prasugrel treatment groupcompared to the c1opidogrel treatment group.

7 A.8. Risk-Benefit Analysis: Bleeding as a Function of Time

Relative to clopidogrel, the principal risk associated with prasugrel is the risk of bleeding, andthe principal benefit is the prevention of non-fatal myocardial infarction. By considering theendpoint events prevented by prasugrel relative to the bleeding events attributed to prasugrel,an actual cumulative benefit-risk ratio can be calculated cumulatively over time. The cumulativepercentage of endpoint events prevented was calculated by subtracting the event rates forprasugrel and c1opidogrel in the Kaplan-Meier analysis for the overall ACS population (i.e., themethod used to generate Figure 7). The same approach was used for bleeding events that metthe regulatory definition of a serious adverse event (SAE), TIMI major, and TIMI major or minorbleeds. For each bleeding category, the cumulative delta percent was calculated over time.Finally, at each time point, the percentage of endpoint events prevented was divided by thepercentage of excess bleeding events. The resulting functions represent the cumulativenumber of endpoint events prevented per excess bleeding event, as a function of time (Figure16).

The general shapes of the relations are similar for all the 3 categories of bleeding events. Thetradeoff between efficacy and bleeding is most favorable around day 12, exhibits a gentle"plateau" through approximately Day 30, and declines through day 80, as the numbers ofattributable bleeding events outpace the number of endpoint events prevented. After day 80,the benefit-risk relation is fairly constant (Figure 16, data shown through Day 180).

Prasugrel Secondary Review, page 49 of77